顧客生命週期價值（CLV）是衡量顧客對企業長期價值的重要指標。本研究以新鮮食品超市為例，利用深度神經網絡（DNN）模型來預測常客的CLV。通過對超市一年的消費數據進行預處理和特徵選擇，我們採用DNN與Elastic Net回歸相結合的模型，並使用移除離群值的方法來提高預測準確性。我們比較了不同預測期間的預測結果，並發現使用兩個月的消費數據來預測未來五個月的CLV效果最好。此外，我們還研究了各個特徵對CLV的重要性，並發現消費金額、月份、消費頻率以及不同類別購買金額比例對CLV有顯著影響。如果將來有更長時間範圍的數據集，我們可以嘗試去除季節因素，以預測超過一年的CLV。;This research paper explores the application of Deep Neural Networks (DNN) in predicting Customer Lifetime Value (CLV) for regular customers in fresh food supermarkets. Real transaction data spanning 11 months is analyzed, and various data preprocessing techniques, outlier handling methods, and algorithms are compared to develop an accurate CLV prediction model. The findings highlight the importance of outlier removal and feature selection, with the combination of DNN and Elastic Net regression demonstrating the best performance. The study also identifies the optimal forecasting period. The research provides practical insights for businesses seeking to improve CLV forecasts and showcases the potential of DNN in CLV prediction across industries. Overall, this work contributes to advancing CLV prediction methodologies and offers a framework for enhancing accuracy in predicting CLV for regular customers in fresh food supermarkets.
